SPEECH EVALUATION SYSTEM

• Main Authors: MATSUOKA HOSEI, SAWAYAMA ATSUKI, TAKEYA KENGO, OKAMOTO NORITAKA, MURAKAMI SOICHIRO
• Format: Patent
• Language: English; Japanese
• Published: 24.06.2022
• Subjects: ACOUSTICS; ADVERTISING; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND,DEAF OR MUTE; CRYPTOGRAPHY; DIAGRAMS; DISPLAY; EDUCATION; EDUCATIONAL OR DEMONSTRATION APPLIANCES; GLOBES; MUSICAL INSTRUMENTS; PHYSICS; PLANETARIA; SEALS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

To provide a speech evaluation system with which it is possible to appropriately detect pronunciation errors in the speech of a speaker.SOLUTION: A speech evaluation system 1 comprises: a conversion unit 33 for acquiring the result of having voice-recognized the speech of a speaker S and converting the voice recognition result into a character string that represents pronunciation; a calculation unit 34 for calculating an editing distance between an assumed character string that represents the pronunciation of a word assumed to appear in the speech of the speaker and one or a plurality of speech character strings included in the character string representing the pronunciation and having been converted by the conversion unit; a detection unit 35 for detecting, as a pronunciation-error character string, a speech character string, among the one or plurality of speech character strings, the editing distance of which is smaller than or equal to a prescribed value and which is not the same as the assumed character string; and an output unit 38 for outputting the pronunciation-error character string. The calculation unit takes into account the pronunciation classification of consonants included in the assumed character string and the speech character string as it calculates the editing distance.SELECTED DRAWING: Figure 2 【課題】話者の発話における発音の誤りを適切に検出することができる発話評価システムを提供すること。【解決手段】発話評価システム１は、話者Ｓの発話を音声認識した結果を取得し、該音声認識した結果を、発音を表す文字列に変換する変換部３３と、話者の発話に出現すると想定される単語の発音を表す文字列である想定文字列と、変換部によって変換された発音を表す文字列に含まれる一又は複数の発話文字列との編集距離を算出する算出部３４と、一又は複数の発話文字列のうち、編集距離が所定値以下であり、且つ、想定文字列と同一の文字列ではない発話文字列を、発音誤り文字列として検出する検出部３５と、発音誤り文字列を出力する出力部３８と、を備え、算出部は、想定文字列及び発話文字列に含まれる子音の発音分類を考慮して、編集距離を算出する。【選択図】図２